search

WayScience / phenotypic_profiling

Machine learning for predicting 15 single-cell phenotypes from cell morphology profiles
Creative Commons Attribution 4.0 International
1 stars 3 forks source link

Use 2015 data & remove holdout set #5

Closed roshankern closed 1 year ago

roshankern commented 1 year ago

This PR is ready for review! This PR incorporates a newer version of mitocheck_data, downloading the 2015 MitoCheck dataset and merging it with the older dataset. The pipeline is then rerun with this expanded dataset.

The "holdout" dataset is also removed, leaving only the training and testing data subsets. Our logic here is that the application of the final phenotypic profiling model to other datasets (ex Cell Health) will validate the model in the same way a holdout dataset would.

roshankern commented 1 year ago

Thank you for the review @d33bs!

Notebook files 2.train_model/train_model.ipynb and 4.interpret_model/interpret_model.ipynb should not have .py files in this PR because their python files had no changes to be tracked in this PR (the Jupyter files were just rerun).

roshankern commented 1 year ago

Accidentally merged this PR without approval, but @d33bs and I discussed that everything was good to merge.